Efficient definitive screening designs
to optimize the freeze-drying process
Olga YeeNCS 2018
Paris, France
Lyophilized products
Examples:
2
Lyophilization
3
Very expensive processIt can take 1 week to finish one lyophilization run.
Lyophilization Tray Template – Sampling Center and Edge Vials
4
Typical Lyophilization Cycle
5
Swamp Time ~15%
Secondary Drying
Primary Drying
Freezing
Design choice
ChallengeDesign a study with 8 factors in less than 20+ runs with minimal risk of a follow-up study. Each lyo run takes one week to complete.
Some design options1. Fractional factorial design: Resolution IV design in 16 runs,
meaning two-factor interactions are completely confounded with other two-factor interactions.
2. Central composite design: prohibitive in terms of number of runs (over 60 runs).
3. Definitive screening design
Advantages of definitive screening designsReference: Jones and Nachtsheim, 2011, Journal of Quality Technology, “A Class of Three-Level Designs for Definitive Screening in the Presence of Second-Order Effects”
Fewer runs: (2m+1) where m is the number of factors. Main effect estimates are unbiased by any second-order effect. Two-factor interactions are not completely confounded with other two-
factor interactions, although they may be correlated. With 6 through (at least) 12 factors, the designs are capable of estimating
all possible full quadratic models involving three or fewer factors with very high levels of statistical efficiency.
DSD with 8 factors in only 20 runs
DoE Parameter Low Middle HighDrug Concentration (mg/mL) 10 30 50
Lyoprotectant (wt%) 6.0 7.5 9.0Primary Drying Tshelf (°C) -13 -8 -3
Chamber Pressure (mTorr) 50 100 150Secondary Drying Duration (hours) 5.0 7.5 10.0Temperature Ramp Rate (°C/min) 0.2 0.6 1.0
Fill Volume (mL) 6.0 7.5 9.0Instrument LyostarII or Virtis
A definitive screening DoE was designed to test the effects of eight process and formulationfactors on many lyophilization responses, including primary drying time and producttemperature.
8
Defining the end of primary drying: Intersection of product temp and shelf temp
Note the difference in orange and blue thermocouples: 6.3 hours.
9
Run 16
TM
& 4
more
-24
-22
-20
-18
-16
-14
-12
-10
-8
-6
50 52 54 56 58 60 62 64 66 7067.8 hrs61.5 hrs
Time (hrs)
TM
TF
BR
BM
Shelf InRun 16 Intersection of shelf temp and
actual temp
SamplePrimary drying
time Shelf TempTM (blue) 67.8 -8
TF 58 -8.1BR 57.8 -8
BM (orange) 61.5 -8.1
Defining the end of primary drying
10
“Product temperature approaching the shelf temperature set point (i.e., “offset” in Fig. 2) is commonly taken as an indication of the end of primary drying.”
S. M. Patel, T. Doen, and M. J. Pikal, AAPS Pharm.Sci.Tech., 11, 2010
Four-parameter logistic curve:lower asymptote cupper asymptote dSlope bEC50, or e; where 50% of the response is expected
Defining the end of primary dryingMathematical Method: Fourth derivative of the 4-PL
11
• Need to prove that offset is reached at the maximum value of 4th derivative over the 2nd portion of the curve.
• After completing all 20 experiments, the difference in model quality between the two methods was not significant.
Variance component structure of the lyophilization data
Between-run variation consists of a fixed and a random part. Within-run variation is due to random variation after accounting for location effects: • tray position (top, bottom)• thermocouple position (front,
middle, rear)• as well as analytical and
sampling variation
12
Mixed Model for Primary Drying Time
13
Variance components Table• Properly accounting for sources of variation leads to a decomposition of variance components into
whole-plot error and split-plot error terms. • Incorrectly pooling these two sources of variation into one leads to a more sporadic significance of
effects that may not be real (inflated type I error rate, biased t-ratios and p-values).
14
Term Estimate Std Error DFDen t Ratio Prob>|t|
Intercept 37.769 0.852 10.414 44.35 <.0001
Shelf.Temp(-13,-3) -5.509 0.572 10.312 -9.63 <.0001
Fill.volume(6,9) 8.320 0.576 10.223 14.45 <.0001
Chamber.Pressure(50,150) -3.171 0.613 9.855 -5.17 0.0004
DS.conc(10,50) 3.355 0.573 10.493 5.85 0.0001
Freezing.rate(0.2,1) 2.455 0.587 10.279 4.18 0.0018
Instrument[1] -0.332 0.497 10.442 -0.67 0.5184
Fill.volume*Shelf.Temp -2.089 0.665 10.624 -3.14 0.0097
Chamber.Pressure*Chamber.Pressure 2.801 1.100 10.192 2.55 0.0287
tc.loc[front] -0.994 0.299 92.461 -3.33 0.0013
tc.loc[middle] 3.749 0.248 92.430 15.09 <.0001
tc.tray[bottom] -0.164 0.190 92.423 -0.86 0.39
Was DSD a good choice?Final model has:• Six main effects: Vial Fill Volume, Shelf Temperature, Drug Substance
Concentration, Chamber Pressure, Freezing Rate, and Instrument• One quadratic effect for chamber pressure• One two-factor interaction: Fill Volume*Shelf Temperature• Location effects within run: tray position and thermocouple locationDefinitive screening design proved to be a success. No follow-up study is needed to further understand and optimize the freeze-drying process. Another monoclonal antibody showed excellent agreement with this model.
15
Conclusions and Future WorkProcess understanding
An eight parameter mAb lyophilization DoE was completed, testing both formulation andprocess variables. The DoE may enable improved selection of formulation and processparameters for new lyophilization candidates and highlights relationships betweenparameters and product/process attributes.
This study can be augmented to expand the design space to a lower shelf temperature, fill volume, instrument type, etc.
Business impact
Significant savings in time and drug substance quantity for delivering drugs for clinicalstudies.
Several other drugs were developed using knowledge from this study.
16
References
B. Jones, C.J. Nachtsheim (2011) “A Class of Three-Level Designs for Definitive Screening in the Presence of Second-Order Effects” Journal of Quality Technology, 43:1, 1-15J. Goldman, H. More, O. Yee et al, “Optimization of Primary Drying in Lyophilization During Early-Phase Drug Development Using a Definitive Screening Design With Formulation and Process Factors” J Pharm Sci 2018 Oct 8; 107(10): 2592-2600
17
Acknowledgements
18
Engineering Technologies Parenterals Science & Technology